jffs v1 - runs out of space...what next?

jffs v1 - runs out of space...what next?

[Craig Gwydir]

For rather evil purposes, we are using the linux 2.4.16 jffs code on a
read-write filesystem.
Occasionally, I use rsync to update this jffs filesystem then reboot.

A problem occurs when the jffs v1 filesystem becomes full.  The next reboot
of the target displays:

jffs_scan_flash(): did not find even a single chuck of freespace. That is
BAD!
jffs_scan_flash(): free sized accounting screwed
jffs_scan_flash(): free_chukc_size1 == 0xf00000, free_chuck_size2 == 0x0,
                   fmc->free_size == 0x0

JFFS: failed to mount root device
kernel panic: Unable to mount root fs
kdb>

There is only about 12MB on my 16MB flash in use tho.  I believe rsync
created a tmp file
during the transfer of a large file which caused useable space to fill.


Question:

1) Any internal JFFS config settings I can tweak to avoid this condition?
   (Yes, I know, the "don't do that" suggestion works here ;)

2) Has anybody had this condition of jffs running out of space, and they
experience the same
   condition?

Yes, I know I probably should run jffs mounts as real-only since it is
flash.
I thought that jffs v1 was supposed to be robust enough to handle failures,
including
"no space on device" errors.

Thanks,
Craig

creating jffs filesystems with mkfs.jffs

[Craig Gwydir]

I am using an Intel E28F128 flash part in my design along with jffs v1 (x86,
linux 2.4).

The Intel flash part is organized as 128 x 128KB blocks.

Question: Are there any options I should use on mkfs.jffs to accomodate
this?
          Should I use the mkfs.jffs -e option and set it to 131072 for the
erase size
          when creating the jffs filesystem?

Thx,
Craig